AMD Readies FX-8370, FX-8370E Microprocessors.

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Games - 80/20 in favor of 2500K
Single threaded apps - 2500K
Highly Multi threaded apps - FX 8370E
Power Consumption - if you really care then 2500K

Conclusion: Don't do it unless you have a board laying around. You would be better served by getting a 4790K

Updated that for you :p
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
Hardware.fr Haswell-E review has an interesting chess game based bench, that is Houdini 4, wich was recently released and wich is quite Haswell friendly since there s 19% higher IPC than previous gen but the most interesting is that there s two exe files, one optimised for AMD and the other for Intel.

Despite said 19% gain for Haswell the FX8350 is slightly ahead of the 4770k by 4%, this tells that once softs are correctly optimised for AMD (H)as well (pun intended) the FX show its strength and real competitivness despite the two year old release in comparison to a more recent competing product.

getgraphimg.php



http://www.hardware.fr/articles/924-14/ia-echecs-stockfish-fritz.html
 
Last edited:

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
I don't see the octal core AMD CPU getting anywhere near the Intel CPU. Must be that weird French math.
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
I don't see the octal core AMD CPU getting anywhere near the Intel CPU. Must be that weird French math.

Denial as argument..??.

Sorry, the number is what it is, you wont be surprised that i have much more faith in Marc Prieur s test than to your uninformed opinion.

Edit : France is the country of mathematicians, so much for the weird maths, you managed to be wrong twice actualy in two sentences...
 
Last edited:
Aug 11, 2008
10,451
642
126
Talk about denial as an argument. Its back to the old "cant be the FX, must be the software" argument. Thought that excuse died out long ago.
 

Enigmoid

Platinum Member
Sep 27, 2012
2,907
31
91
Denial as argument..??.

Sorry, the number is what it is, you wont be surprised that i have much more faith in Marc Prieur s test than to your uninformed opinion.

Edit : France is the country of mathematicians, so much for the weird maths, you managed to be wrong twice actualy in two sentences...

I don't know why you are using a single benchmark to make a conclusion. The FX series has its strong points (encoding, compression, integer math). It has its weak points. One benchmark doesn't prove anything. Most synthetic integer benchmarks favour the FX just as most synthetic FP benchmarks favour HW.

Big deal.

7z-decomp.png
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
While we re on chess games the other test use Stockfish 5 wich according to the reviewer is GCC 4.8 compiled in 3 versions,a classical one (SSE2 i m assuming), a SSE4 version wich bring 3% improvement for AMD and Intel CPUs and a third version using BMI instructions for Haswell wich add another 2 %, Hardware.fr used the most efficientsversions for each CPU and the results are about the same as Houdini 4.

getgraphimg.php


http://www.hardware.fr/articles/924-14/ia-echecs-stockfish-fritz.html
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Denial as argument..??.

Sorry, the number is what it is, you wont be surprised that i have much more faith in Marc Prieur s test than to your uninformed opinion.

Edit : France is the country of mathematicians, so much for the weird maths, you managed to be wrong twice actualy in two sentences...

AMD 14,452 - Intel 19,458. You were saying FX is how much faster than Haswell?
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
I don't know why you are using a single benchmark to make a conclusion. The FX series has its strong points (encoding, compression, integer math). It has its weak points. One benchmark doesn't prove anything. Most synthetic integer benchmarks favour the FX just as most synthetic FP benchmarks favour HW.

Big deal.

It was in respect of equivalent optimisation for both CPUs, most benches are optimised for Intel pipeline, i dont know about 7zip since it performs quite well with the FX but if you want to see the results of a FP rigged bench check Anandtech reviews and 3D particle movement wich show 50% higher IPC for Bay trail compared to Kabini while we all know that in FP the latter has 30% better IPC, this bias extend of course to the FX wich is considerably slowed for the purpose.
 

Enigmoid

Platinum Member
Sep 27, 2012
2,907
31
91
While we re on chess games the other test use Stockfish 5 wich according to the reviewer is GCC 4.8 compiled in 3 versions,a classical one (SSE2 i m assuming), a SSE4 version wich bring 3% improvement for AMD and Intel CPUs and a third version using BMI instructions for Haswell wich add another 2 %, Hardware.fr used the most efficientsversions for each CPU and the results are about the same as Houdini 4.

getgraphimg.php


http://www.hardware.fr/articles/924-14/ia-echecs-stockfish-fritz.html

Yep and you notice two things. First is that HW isn't massively better than IVB. Second is that the +4% turns into -6%.

The isn't any conspiracy here.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Talk about denial as an argument. Its back to the old "cant be the FX, must be the software" argument. Thought that excuse died out long ago.

Nope, it will be around forever. Just recently the subject of Intel compilers was brought up again.

But then, look at my new sig, it will explain a lot.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
It was in respect of equivalent optimisation for both CPUs, most benches are optimised for Intel pipeline, i dont know about 7zip since it performs quite well with the FX but if you want to see the results of a FP rigged bench check Anandtech reviews and 3D particle movement wich show 50% higher IPC for Bay trail compared to Kabini while we all know that in FP the latter has 30% better IPC, this bias extend of course to the FX wich is considerably slowed for the purpose.

Lol. Rigged benchmarks. It's all coming together, you're just a conspiracy theorist.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
....no comment....

Ok, so it takes 8 AMD cores to match 4 Intel cores (when comparing to an old Intel CPU and not the latest highest performing). We knew this a long time ago, but thanks for pointing it out again.
 

Enigmoid

Platinum Member
Sep 27, 2012
2,907
31
91
It was in respect of equivalent optimisation for both CPUs, most benches are optimised for Intel pipeline, i dont know about 7zip since it performs quite well with the FX but if you want to see the results of a FP rigged bench check Anandtech reviews and 3D particle movement wich show 50% higher IPC for Bay trail compared to Kabini while we all know that in FP the latter has 30% better IPC, this bias extend of course to the FX wich is considerably slowed for the purpose.

Not sure if serious.

66506.png


The J1900 is running at 2.42 ghz and gets 65 points. 26.9 points/Ghz

The 5350 is running at 2.05 ghz and gets 53 points. 25.8 points/Ghz

Pretty much the same. You are off in your claim by more than an order of magnitude.

FX and Kabini are completely different when it comes to FP calculations. Kabini is quite strong while FX is weak.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Not sure if serious.

66506.png


The J1900 is running at 2.42 ghz and gets 65 points. 26.9 points/Ghz

The 5350 is running at 2.05 ghz and gets 53 points. 25.8 points/Ghz

Pretty much the same. You are off in your claim by more than an order of magnitude.

FX and Kabini are completely different when it comes to FP calculations. Kabini is quite strong while FX is weak.

Oops, you posted facts.
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
Lol. Rigged benchmarks. It's all coming together, you're just a conspiracy theorist.


Ad hominem and thread crapping because you have no numbers and absolutely no arguments, but anyway let s see the numbers in FP :

CB 11.5

http://uk.hardware.info/reviews/533...ap-desktop-platforms-benchmarks-cinebench-115

Kabini 2.05 is 11% faster than a BT at 2.4 , 30% better IPC than J1900.

POV ray

http://www.computerbase.de/2014-04/...i-sockel-fs1b-test/3/#diagramm-pov-ray-37-rc7

Kabini has rougly 35-40% better IPC than the J1900.


Now let s look at Anandtech rigged bench :

64166.png


64167.png


What happened.??..

And of course the reviewer conclusion is :

The Intel CPU takes the crown in floating point tests,

I did read that Anand did quit the publication staff and that Ryan Smith is the new publication director, hope Anand get more time and post in the forum, i m really curious how they managed to draw such a rigged conclusion, actualy it s the only site that say so.

My opinion of this bench is that it s actualy a memory bandwith test as it s about the only sector where Bay trail has an advantage over Kabini, check Haswell-e review for the FX scores.
 

crashtech

Lifer
Jan 4, 2013
10,695
2,294
146
It's unseemly and off-putting to call results you disagree with "rigged." Either they made a mistake on the benches, or you are just flat wrong. To accuse Anandtech of rigging a benchmark is laughable.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,587
10,225
126
It's unseemly and off-putting to call results you disagree with "rigged." Either they made a mistake on the benches, or you are just flat wrong. To accuse Anandtech of rigging a benchmark is laughable.

I don't think he's necessary saying that AT rigged the benches themselves. It might have been down to whomever compiled the binaries that they used. Of couse, if they compiled them all themselves from source, then... maybe.
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
Pretty much the same. You are off in your claim by more than an order of magnitude.

You re right , i confused 50% better IPC with 50% artificialy improved FP IPC for BT but still that s considerable, if we take an average of CB11.5 and Povray Anandtech s 3D particle get a result that improve BT IPC by almost 50% to get it 5-10% faster than a Kabini.

Same bench will show the same oddity with the FX when compared to other FP benches, the FX scores are identicaly reduced in respect of the i5 and i7 (instructions have nothing to do with it since Kabini has more recent FP instructions than BT so the discretanpcy is due to the soft).
 

T_Yamamoto

Lifer
Jul 6, 2011
15,007
795
126
The forums is not connected to Anandtech, it's just loosely affiliated.

I'm not sure how those reviews are rigged. They look fine to me.
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
It's unseemly and off-putting to call results you disagree with "rigged." Either they made a mistake on the benches, or you are just flat wrong. To accuse Anandtech of rigging a benchmark is laughable.

They are professionals, how could they not know what amateurs know ..??.

One could only be surprised, unless of course that they are ignorant of what is published on other sites, also how could they come to this conclusion based on a single bench that nobody knows .?

Would be interesting to know who is the editor/conceptor of this bench and what it does actualy, as said my opinion is that it s a memory bandwith bench that is not representative of actual FP perfs.
 
Last edited:

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
FX 8370 thread turned into an Intel cheating thread :confused:

Same players, same agenda.
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
FX 8370 thread turned into an Intel cheating thread :confused:

Same players, same agenda.

You couldnt resist doing ad hominem of course and deflection since nobody, set apart you, said that intel has something to do with this test.

This bench was used with the FX as well so it s perfectly relevant with this thread, moreover because it has the same effect on the FX that on Kabini when compared to other FP benches.

That was my last post directed to you, i m not interested by useless polemics but by the technical aspect of thing, first time that i use the ignore button in any site.